Strategic Role of Green Orientation Environment in Shaping Employee Green Behavior: A Study of State-Owned Energy Enterprises in Indonesia

It is widely recognized that economic development often has a negative correlation with environmental conditions. Today, not only governments but also the public are increasingly aware of the environmental degradation that can accompany economic growth [1]. Environmental activists have long urged both industrial companies and governments to adopt "go green" initiatives. The industrial revolution led to the massive exploitation of energy resources, which in turn contributed to serious environmental damage. In response, the Indonesian government has required all companies to improve their environmental practices [2].

To reinforce this effort, the government enacted Government Regulation (PP) No. 22 of 2021 concerning the Implementation of Environmental Protection and Management. Additionally, the Ministry of Environment and Forestry issued Regulation No. 3 of 2021, which outlines standards for business activities under risk-based licensing in the environmental and forestry sectors. These regulations were issued in light of persistent environmental damage across Indonesia. Prior to these governmental efforts, the Ministry of State-Owned Enterprises had already mandated corporate environmental responsibility through Regulation No. PER-01/MBU/2011 concerning the Implementation of Good Corporate Governance in State-Owned Enterprises. This regulation instructs all directors to ensure environmental sustainability in all operational areas of state-owned enterprises (SOEs) [3]. As a result, corporate management has increasingly been driven to implement environmentally responsible practices. This implies that all companies, particularly those engaged in production, must prioritize environmental stewardship. Environmentally friendly production begins with adopting a "Go Green" philosophy [4].

PT Perusahaan Gas Negara Tbk, a state-owned enterprise engaged in natural gas production, is equally obligated to integrate the "Go Green" philosophy as a tangible commitment to sustainability. While the company operates branches across Indonesia, two branches Medan and Dumai are suspected of not fully adopting employee green behavior. As a large enterprise with a nationwide presence, it is the responsibility of management to ensure that all employees uphold a shared commitment to environmental sustainability. To provide an initial overview of employee green behavior, a preliminary survey was conducted between October 2 and October 31, 2023. The results are presented below.

Figure 1 shows that Employee Green Behavior (EGB) at PT Perusahaan Gas Negara, Tbk, particularly at the Medan and Dumai branches remains suboptimal. This situation poses a serious risk to the company’s ability to sustain its operations, as failure to support environmental sustainability may lead to long-term reputational and regulatory consequences. The company could face environmental disputes with communities, environmental NGOs, and even local governments that perceive the company as negligent in preserving the environment.

Perusahaan gas negara, tbk medan branch & dumai branch

3.1 Demographic characteristics of respondents

To provide a deeper contextual understanding of the study population, this section presents the demographic profile of the respondents drawn from two operational branches of PT Perusahaan Gas Negara, Tbk, located in Medan and Dumai. The demographic data include four key variables: gender, working period, age, and educational level. These attributes are essential to consider, as they may shape the respondents’ awareness, orientation, and behavioral responses related to environmental sustainability and organizational practices.

Understanding the demographic structure also enables identification of potential structural or cultural differences between branches that could influence employee engagement with green initiatives. Furthermore, variations in these characteristics may provide insights into how GIM and GT are perceived and internalized across different respondent segments. Table 1 summarizes the distribution of respondents by demographic categories.

Table 1. Demographic characteristics of respondents

Before proceeding to inferential analysis, it is essential to assess the distributional properties of the primary research constructs through descriptive statistics. This step provides a foundational understanding of central tendencies and variability in respondent perceptions, while also supporting an initial review of normality assumptions. The following key indicators are evaluated for each construct: number of observations (N), minimum and maximum values, mean, standard deviation (SD), skewness, and kurtosis. These measures serve as the basis for ensuring that the data are suitable for Partial Least Squares Structural Equation Modeling (PLS-SEM) and for identifying any initial trends in perceptions of GIM, GT, GOE, and EGB. A summary of the descriptive statistics for each construct is presented in Table 2.

Table 2. Demographic characteristics of respondents

As presented in Table 2, each construct was assessed using data from 230 respondents, with all variables exhibiting a minimum value of 2 and a maximum value of 5. This distribution suggests the absence of extreme outliers and confirms that responses remained within the expected range of the Likert scale. Among the four constructs, GT recorded the highest mean (M=3.62), indicating a generally positive perception among employees toward the environmental training initiatives implemented by the organization. Conversely, GOE had the lowest mean (M=3.35), suggesting that while training is actively conducted, the internalization and personal adoption of green values may still require reinforcement.

The standard deviations ranged from 0.66 to 0.76, reflecting moderate variability in responses across all constructs. The lowest variability was observed in EGB (SD=0.66), implying a relatively consistent perception among respondents regarding green behavioral practices in their workplace. Furthermore, the skewness and kurtosis values for all variables fall well within the acceptable range of ±2, indicating that the data is approximately normally distributed. This supports the suitability of the dataset for parametric analysis, such as Structural Equation Modeling using PLS-SEM. These statistical indicators collectively demonstrate that the dataset meets the assumptions for advanced modeling and is robust enough for subsequent hypothesis testing.

3.2 Measurement model estimation

The evaluation of the measurement model in this study employed a second-order factor approach, consistent with the hierarchical nature of the theoretical constructs. Each variable, including GIM, GT, GOE, and EGB, was conceptualized as a higher-order latent construct, composed of several first-order dimensions and their corresponding indicators.

The embedded two-stage approach was applied, wherein the first stage validates the indicator dimension relationship (first order), followed by the second stage evaluating the dimension construct relationship (second order).

In Stage 1, the outer loadings of all indicators within their respective dimensions were found to exceed 0.70, indicating strong indicator reliability. Additionally, the composite reliability (CR) scores for each dimension were all above the threshold of 0.70, demonstrating internal consistency reliability.

Furthermore, the Average Variance Extracted (AVE) values for each dimension surpassed 0.50, supporting good convergent validity. Lastly, the Heterotrait-Monotrait Ratio (HTMT) values were all below the accepted cutoff of 0.90, confirming the discriminant validity between dimensions. These results affirm the robustness of the measurement model and the adequacy of the reflective indicators used in capturing the latent constructs, as summarized in Table 3.

As shown in Table 3, all HTMT values between pairs of dimensions are below the critical threshold of 0.90, thereby confirming the discriminant validity of the constructs in the first-order measurement model. Among the notable values, the HTMT score between In-Role Green Behavior and Extra-Role Green Behavior stands at 0.895. Although relatively high, it remains within the acceptable range, indicating no issue of multicollinearity. Likewise, theoretically related constructs such as Awareness and Responsibility (HTMT=0.887) and Use of Technology and Methodology (HTMT=0.865) demonstrate strong conceptual connections while maintaining distinct statistical identities. These findings reinforce the discriminant strength and structural soundness of the measurement model, ensuring that each dimension contributes uniquely to its respective latent construct.

Table 3. Discriminant validity-HTMT stage first

Table 4. Loading factor, composite reliability, AVE

To further evaluate the reliability and convergent validity of the second-order constructs, three standard PLS-SEM criteria were assessed: factor loadings, CR, and AVE. Factor loadings above 0.70 indicate robust indicator reliability, while CR values exceeding 0.70 confirm satisfactory internal consistency. Similarly, AVE values greater than 0.50 ensure adequate convergent validity. All second-order constructs in this study meet these thresholds, establishing a reliable and valid measurement foundation for structural model analysis. The detailed results are presented in Table 4.

As shown in Table 4, all loading factor values for the respective dimensions exceed the recommended threshold of 0.70, indicating that the observed indicators reliably represent their associated latent constructs. For instance, the loading for Extra-Role Green Behavior under the EGB construct is particularly high at 0.920, reflecting strong indicator relevance. All constructs exhibit CR scores above 0.90 specifically, GIM (0.912), GT (0.900), GOE (0.908), and EGB (0.902) which demonstrates excellent internal consistency. Similarly, the AVE values for all constructs surpass the minimum criterion of 0.50, confirming convergent validity by explaining more than half of the variance in their respective indicators. Notably, EGB records the highest AVE at 0.821, reinforcing the strong construct validity of this latent variable. Collectively, these findings establish the reliability and validity of the measurement model, ensuring that each construct is measured both accurately and consistently. This robust measurement foundation enables reliable interpretation of the structural model in the next stage of analysis. To assess discriminant validity at the construct level, the HTMT of correlations was calculated for the four second-order latent variables. HTMT values below 0.90 confirm that constructs are statistically distinct and do not overlap conceptually. The results are summarized in Table 5.

As presented in Table 5, the HTMT values among all second-order latent constructs remain well below the recommended threshold of 0.90, thereby establishing discriminant validity at the construct level. The highest inter-construct correlation is between GIM and GT (HTMT=0.678), which is theoretically plausible given their shared function as internal organizational enablers. The HTMT value between EGB and GOE is 0.556, indicating a moderately strong relationship that supports the mediating role of GOE. Additionally, GT and EGB are related with an HTMT of 0.611, affirming the hypothesized influence of training on green workplace practices. These results demonstrate that each construct is conceptually and statistically distinct, reinforcing the structural model’s empirical soundness and validity. To assess the overall model fit and predictive relevance, several quality criteria were employed:

1. R-Squared (R²) values indicate the explained variance of endogenous constructs. The R² for EGB is 0.326, while GOE has an R² of 0.245, both of which fall within the moderate explanatory power category [33]. This suggests that the exogenous variables provide meaningful, though not exhaustive, explanation for the target constructs.
2. Q-Squared (Q²) values, which evaluate predictive relevance, are 0.254 for EGB and 0.169 for GOE. These values exceed the minimum threshold of zero, indicating moderate predictive relevance [34].
3. Standardized Root Mean Square Residual (SRMR) was calculated to assess model fit. The SRMR value of 0.071 falls below the critical cutoff of 0.08 [35], indicating an acceptable model fit between the empirical and predicted correlation matrices.

Table 5. Discriminant validity

Table 6. Model quality criteria

Together, these indicators provide robust evidence that the model not only fits the data well but also possesses reasonable explanatory and predictive power. To further evaluate the model’s out-of-sample predictive capability, the PLS Predict procedure was employed. The results are detailed in Table 6 and discussed in the following section.

As illustrated in Table 6, the results of the PLS Predict analysis offer a comprehensive evaluation of the model’s out-of-sample predictive power. Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) values are compared between the PLS-SEM model and a naïve linear regression (LM) benchmark for each of the key indicators under the endogenous constructs. In the case of Extra-Role Green Behavior, the LM model yields slightly lower RMSE (0.863 vs. 0.876) and MAE (0.699 vs. 0.690), indicating marginally better predictive accuracy. Similarly, for In-Role Green Behavior, the PLS model produces a marginally higher RMSE (0.915) and MAE (0.736) than LM, suggesting slightly inferior performance on this indicator.

However, for the majority of the other indicators including Compliance, Collaborate, Awareness, and Responsibility the PLS model performs consistently and competitively, with MAE values that are either lower or very close to the LM benchmark. Notably, PLS achieves better MAE on 'Responsibility' (0.734) and matches LM performance on 'Collaborate' (0.704). Although RMSE values for PLS are slightly higher in some cases, they remain within acceptable margins, and the overall predictive errors are not excessive.

These findings collectively indicate that while the LM model performs marginally better in a few specific cases, the PLS model maintains robust and reliable predictive relevance across the board. This result justifies the selection of PLS-SEM, particularly given its ability to handle complex models, account for mediation effects, and provide meaningful insights even with modest sample sizes. The PLS Predict analysis thus reinforces the external validity and generalizability of the proposed model in predicting green behavioral outcomes in organizational contexts.

As shown in Table 7, the path coefficient from GIM to EGB is not statistically significant (β=0.091, p=0.165), suggesting that direct managerial involvement alone does not substantially influence employees’ green behavior. However, GIM has a significant effect on GOE (β=0.210, p=0.002), which in turn significantly predicts EGB (β=0.274, p<0.001). This confirms the indirect effect of GIM via GOE (β=0.058, p=0.007), establishing a statistically significant mediating pathway.

Table 7. Parameter estimates and mediating effect

Similarly, GT exerts a strong direct influence on both GOE (β=0.338, p<0.001) and EGB (β=0.323, p<0.001). It also demonstrates a significant indirect effect through GOE (β=0.093, p=0.002). These results validate the proposed mediation model and emphasize the strategic role of GOE in translating organizational initiatives into observable behavioral outcomes. In summary, while the direct effect of GIM on EGB is limited, its indirect influence through environmental orientation is pivotal. Conversely, GT exerts both direct and mediated effects, reinforcing employee awareness and commitment to environmentally responsible practices. These findings underscore the importance of cultivating a green-oriented culture as a critical intermediary in achieving sustainable workplace behavior.

3.3 Discussions

The findings of this study reveal nuanced insights regarding the mechanisms through which organizational strategies shape environmentally responsible employee behavior. Notably, GIM did not demonstrate a significant direct influence on EGB (β=0.091, p>0.05), contradicting previous studies that positioned managerial involvement as a primary driver of pro-environmental conduct in the workplace [21, 36]. This suggests that managerial efforts, when not internalized or operationalized at the employee level, may have limited behavioral impact. However, GIM exerted a significant indirect effect on green behavior via GOE (indirect β=0.058, p<0.01), highlighting the importance of shaping environmental values and awareness as a mediating mechanism. This supports the assertion that managerial influence is most effective when it fosters an internalized orientation toward sustainability rather than merely enforcing top-down initiatives [19].

The Ecocentric Management Theory emphasizes the dual responsibility of organizations to achieve profitability and to maintain environmental sustainability. In line with this principle, the study confirms that employees with strong environmental orientation tend to exhibit higher levels of green behavior (β=0.274, p<0.001). This reinforces the role of internal values and awareness in mediating the relationship between organizational strategies and individual actions. Employees who perceive environmental protection as part of their professional identity are more likely to translate that orientation into tangible eco-friendly practices both within the workplace and in their surrounding community. Equally important, the role of GT emerged as both directly and indirectly influential. The training significantly enhanced both GOE (β=0.338, p<0.001) and EGB (β=0.323, p<0.001), and its indirect pathway through orientation was also statistically significant (β=0.093, p<0.01). These findings align with previous literature [1, 7, 15, 16], validating the critical function of training as a strategic tool in GHRM frameworks [24]. Training equips employees with knowledge, reshapes attitudes, and empowers behavior change.

From a practical standpoint, the study suggests that management should not solely rely on policy mandates or managerial supervision to instill green behavior. Instead, a more sustainable approach lies in embedding environmental orientation through training and cultural reinforcement. Employees are more likely to adopt consistent and voluntary green behaviors when they understand the purpose behind environmental initiatives and align those values with their personal and professional goals. While direct management involvement may have limited behavioral influence, its capacity to shape orientation remains vital. Moreover, GT serves as a powerful lever providing both knowledge and motivation to foster a culture of environmental stewardship within organizations. This study provides empirical evidence for the mediated pathways by which organizational strategies can be translated into behavioral outcomes, underscoring the importance of environmental orientation as a central construct in driving sustainable workplace transformation.